{
  "id": 9620,
  "title": "Data preprocessed?",
  "url": "/competitions/seizure-detection/discussion/9620",
  "author_name": "",
  "post_date": "2014-06-30T15:47:59.527Z",
  "votes": null,
  "comment_count": 2,
  "views": 1529,
  "content": "<p>Could you enlighten us, as regards the preprocessing that has been applied on the data (if any), (e.g: noise reduction etc) ?</p>",
  "messages": [
    {
      "id": "49866",
      "postDate": "06/30/2014 15:47:59",
      "content": "<p>Could you enlighten us, as regards the preprocessing that has been applied on the data (if any), (e.g: noise reduction etc) ?</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "50057",
      "postDate": "07/04/2014 23:42:56",
      "content": "<p>Here's an FFT of the signal from one channel for Patient 2.</p>\n\n<p>There are two flat bands: one at high frequency (&gt;1000 Hz) and one between 400 Hz and 1000 Hz. My guess is that the &gt;1000 Hz data is artificial random noise (or &quot;fudged&quot;, as they say in second-year lab....)</p>\n\n<p><img src=\"http://small-yellow-duck.github.io/seizure_detection/fft_patient2_ictal3_channel3.png\" alt=\"fft\" width=\"600\" height=\"450\"></p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "50308",
      "postDate": "07/10/2014 18:25:27",
      "content": "<p>This is a very fair question - none of the data has been &quot;fudged&quot;, or artificially generated but I see why you might think so from the frequency spectrum. For the data sampled at 5kHz, a low pass filter with a sharp cutoff at 1 kHz has been applied to avoid aliasing. While the Nyquist theorem (http://en.wikipedia.org/wiki/Nyquist_frequency) requires a minimum sampling frequency 2x the highest frequency in the signal, in our work we typically keep the sampling frequency at 5x the highest frequency.</p>",
      "rawMarkdown": "",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 50057,
      "author_name": "smallyellowduck",
      "author_url": "",
      "post_date": "07/04/2014 23:42:56",
      "content": "<p>Here's an FFT of the signal from one channel for Patient 2.</p>\n\n<p>There are two flat bands: one at high frequency (&gt;1000 Hz) and one between 400 Hz and 1000 Hz. My guess is that the &gt;1000 Hz data is artificial random noise (or &quot;fudged&quot;, as they say in second-year lab....)</p>\n\n<p><img src=\"http://small-yellow-duck.github.io/seizure_detection/fft_patient2_ictal3_channel3.png\" alt=\"fft\" width=\"600\" height=\"450\"></p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 50308,
      "author_name": "bbrinkm",
      "author_url": "",
      "post_date": "07/10/2014 18:25:27",
      "content": "<p>This is a very fair question - none of the data has been &quot;fudged&quot;, or artificially generated but I see why you might think so from the frequency spectrum. For the data sampled at 5kHz, a low pass filter with a sharp cutoff at 1 kHz has been applied to avoid aliasing. While the Nyquist theorem (http://en.wikipedia.org/wiki/Nyquist_frequency) requires a minimum sampling frequency 2x the highest frequency in the signal, in our work we typically keep the sampling frequency at 5x the highest frequency.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "49866": "",
    "50057": "",
    "50308": ""
  },
  "source": "meta"
}